Bo Zhou issues

Results 26 issues of


                                            Bo Zhou

Is it a bug?

https://github.com/PaddlePaddle/PARL/blob/846ebc97fff4e8bcbd9226c82222bef989c139d8/examples/A2C/train.py#L88 The actor runs with the `no wait` mode. Should we wait here until all the weights are set properly?

【xparl/bug】The master node will collapse while checking the cluster status.

## How to reproduce the bug: 1. start the master node. 2. add a new worker to the cluster. 3. check the cluster status by calling `xparl status` ## Why...

xparl

log improvement: telling the user that it's the error occurred at the remote machine.

This setting may deteriorate the torch speed in CPU environment

https://github.com/PaddlePaddle/PARL/blob/develop/parl/core/torch/agent.py#L26

【core】add unit test for benchmark frameworks such as torch

teamcity

【xparl】telling user that the log files occupy large disk space.

XPARL will not clean the log files after the tasks are finished. The users should be notified if the log files occupy large disk space. I suggest that we can...

xparl

[documentation] tutorial of using xparl in docker

[xparl] Initialization problem

How to reproduce the bug? 1. start a task that keeps requesting CPU resources from the cluster. 2. launch the cluster with `xparl start ...`

xparl

print the current back-end framework if two frameworks are found

![image](https://user-images.githubusercontent.com/25046619/103200600-bc345d80-4928-11eb-9cc8-a4979f467bff.png)

[benchmark] add unit tests for torch

Note: The algorithm_base in torch should reuse the functions: `get_weights` and `set_weights`, but we rewrite their implementations at the moment.

eager mode